Discussion

WITHOUT RUNOFF

III.4.5. Discussion

Modern and past hydrology

These results show that runoff in the Bay of Bengal is essential not only for the water dilution but also for the d¹⁸O - salinity relationship, contrary to the Arabian Sea where only atmospheric fluxes are needed to match the observations. However a better understanding of the d¹⁸O - salinity relationship is needed if we want to predict it in the past. So far, studies focusing on this relationship are based on two assumptions. First, this relationship arises from a mixing line between two ÔpureÕ end-members, an oceanic and a freshwater ones; secondly, the freshwater component consists of continental runoff. This latter assumption probably originates in the depleted characteristics of the observed relationship intercept, readily related to the depleted runoff especially in high latitudes (eg Fairbanks, 1982). Given these assumptions, estimation of past runoff characteristics has been proposed to predict a past d¹⁸O - salinity relationship (Rohling and Bigg, 1998, SchŠfer-Neth, 1998). However, as stressed by Craig and Gordon (1965), the latter assumption concerning the freshwater end-member is generally not valid.

50°E 60°E 70°E 80°E 90°E 100°E Longitude

-10°

0°

10°

20°

30°

Latitude

50°E 60°E 70°E 80°E 90°E 100°E

-10°

0°

10°

20°

Latitude

Figure III.21. Spatial distribution of d¹⁸O simulated in the northern Indian ocean.

Top: simulation with runoff. Bottom: difference between without and with runoff simulation.

The 2-box model approximation allows to estimate the d¹⁸O - salinity relationship as:

d¹⁸O= d0 -I S₀ -0

*S+I (3)

where I being the intercept in the d¹⁸O - salinity diagram corresponds to the isotopic composition of the freshwater end-member, actually a combination of all atmospheric fluxes:

I P P R R E E P R E

= + -

+ -

*d *d *d

(4).

The negative sign for E is due to the fact that the evaporation enriches the surface, contrary to P and R. Of course such a box-model is far from the real world, but the present study shows it is indeed a good approximation of it. Equation 3 shows that the ÔpureÕ oceanic end-member corresponds to waters below the mixing layer, which buffer the atmospheric forcing. The ÔpureÕ freshwater end-member consists in all atmospheric terms including the evaporation (Equation 4). Consequently, the intercept I is generally different from dR since in the open ocean the dilution effect R from the continental runoff is not significant compared to P-E The same holds true for R.dR compared to P.dP-E.dE (see discussion by Craig and Gordon, 1965). For the whole open oceanic surface, the following mean values: dE= -4 permil and dP=

-2.5 permil (since dR is more depleted) , E= 1 m/yr and P= 0.9 m/yr, leads to an intercept I=

-17.5 permil, quite comparable to the observed one from the global GEOSECS measurements.

By contrast, along a coast where the runoff dilution R may be important compared to P- E, the intercept I would be closer to dR. This extreme case has been realistically simulated in the Bay of Bengal, as schematized in Figure III.22. Note that d¹⁸O and salinity values higher than d₀ and S₀ are predicted by the model (P+R-E<0 leads to S>S₀ from Equation 1). Without continental runoff, a low intercept I (-13.4 permil) defines a high d¹⁸O - salinity slope (0.39).

Adding the continental runoff R increases the intercept I (which decreases the d¹⁸O - salinity slope), as simulated with both the box and global models. Quantitatively, a runoff approximately equivalent to the precipitation is required in order to get a slope and a intercept similar to the observed ones. This required runoff flux (10 mm/day) represents a dilution of the total runoff (2300 km³/year prescribed in the models, 0.07 Sv) spread over half the Bay of Bengal (~10⁶ km²). In the Arabian Sea, adding the runoff only slightly lowers the slope (from 0.29 to 0.27 with the GCM) because the dilution effect is not significant compared to P-E.

The schematic diagram of Figure III.22a stresses the difference between the intercept I and the runoff dR with the box model. The same difference holds true with the GCM. Both in the Bay of bengal and in the Arabian Sea the prescribed runoff isotopic compositions are rather homogeneous, due to the low spatial resolution of the GISS model. Both basins receive a runoff with a weighted average composition around -4.5 permil, and a limited scatter between the rivers ranging from -4 to -5 permil. This runoff composition is significantly different from the intercepts I of -7.3 permil in the Bay of Bengal and -8.9 permil in the Arabian Sea simulated with the GCM (cf Table III.3).

-14 -12 -10 -8 -6 -4 -2 0 2

d

O (‰ /SMOW)

40 35

30 25

20 15

10 5

0 Salinity

Evaporation (E= -7 mm/day) Runoff

(R= 10 mm/day) Precipitation (P= 11 mm/day)

runoff addition

(S₀ , d₀)

d

¹⁸O = .21*sal - 7.

d

¹⁸O = .39*sal - 13.4

a

-1

d

18 O (‰ /SMOW)

39 38

37 36

35 34

Salinity

simulated modern relationship (d¹⁸O=.21*sal-7.)

Dd_glob~ 1‰

DS_glob~ 1‰

DS assumed LGM relationship (d¹⁸O=.21*sal-6.2)

b

Figure III.22. Schematic diagrams of the present-day d¹⁸O - salinity relationship for the Bay of Bengal, based on the box model concept as a mixing line determined by the ÔpureÕ end-members, an oceanic and a freshwater ones.

a

. The freshwater source derives from the combination of the precipitation and evaporation isotopic fluxes (stippled line). Addition of the runoff (full line) has an important effect on this source composition and thus on the d¹⁸O - salinity slope. Atmospheric fluxes are derived from the GISS GCM. The ÔpureÕ oceanic source comes from the study of Juillet-Leclerc et al (1997).

b.

Implicit

r

elationship for the Last Glacial Maximum when using the modern spatial slope instead of the temporal one: the local

D

S and

D

d variations are assumed to follow the modern spatial relationship.

In the real world, these water and isotopic fluxes may be somewhat different. Runoff composition can be estimated from the few available measurements. In the Bay of Bengal, the Ganges-Bramahpoutra complex shows a seasonal variation between -6 permil in winter and -10 permil after the monsoon season, with an annual weighted average around -9±1 permil (Ramesh and Sarin, 1992; C. France-Lanord (CRPG, France), pers. com. 1997). This depletion associated with the monsoon is due to the contribution of the Himalayan part of the drainage basin which receives depleted water moisture, as well as to the so-called 'amount effect' i.e. a correlation between precipitation amount and isotopic depletion (Dansgaard, 1964).

In the Arabian Sea, the only reported runoff composition is from Mook (1982), cited by Rohling and Bigg (1998), showing depleted values between -11.5 and -9 permil.

Observations of the isotopic content of precipitation have been compiled by the GNIP (Global Network for Isotopes in Precipitation) program of the IAEA/WMO (World Meteorological Organization) (IAEA, 1969-1983). Precipitation over the Bay of Bengal shows an annual value around -4.5 permil, with a weak seasonal deviation (± 1 permil) , and is slightly enriched over the Arabian Sea (-3.5 permil annually, ± 1 permil) (R. Mathieu et al., 'Isotopes in precipitation : a global view', Univ. of Colorado, manuscript in preparation, 1999). Observed isotopic content of evaporation is not attainable since it would require, among others, the isotopic content of the vapour in the boundary layer.

Thus the water and isotopic fluxes derived from the GISS model and prescribed to the oceanic models, as well as the ones used in Figure III.22, are somewhat different from the real ones: water fluxes (P, E and R) seem too large (cf part 2. for the observed fluxes) and their isotopic content is too enriched (dR) or too depleted (dP). However the simulation of realistic d¹⁸O values as well as d¹⁸O - salinity relationship confirms that the consistency of these fluxes is more important than their absolute values. Indeed, the first-order consistency of these fluxes simulated by the atmospheric GCM is good enough to adequately drive the ocean surface composition. But this is far from perfect, as stressed for instance by the shift between observed and simulated intercepts in the Bay of Bengal.

Consequence of climatic change

As the box model is capable of capturing the basic patterns of the d¹⁸O - salinity relationship for the modern climate, we try to estimate possible climatic change of this relationship. We apply the GISS model atmospheric fluxes simulated with Last Glacial Maximum (LGM) boundary conditions (Jouzel et al., 1994) to the box model. In both basins, the simulated evaporation and precipitation water fluxes decrease, especially the precipitation in the Bay of Bengal due to a drop of the monsoonal intensity.

In the Arabian Sea, the LGM slope is close to the modern observed one, and slightly higher than the modern simulated one (cf Table III.3). In the Bay of Bengal, the slope is comparable to the modern one without runoff. However, in this latter basin, the isotopic fluxes -E.dE and P.dP compensate each other, thus the global intercept I could strongly depend on the

runoff composition. The LGM runoff composition is unlikely to be very different from the present one : simulated changes (Jouzel et al., 1994) in the precipitation composition over the Indian drainage basin are small and in opposite directions: a decrease of d¹⁸O over the colder Himalayan region balances an increase of d¹⁸O over the former monsoonal region due to the precipitation decrease. Thus assuming a global intercept I_LGM between -4 and -5 permil, the slope of the d¹⁸O - salinity relationship would have been between 0.15 and 0.20, even accouting for the d¹⁸O and salinity enrichments of the ÔpureÕ oceanic end-member (by 1 permil). If so, the d¹⁸O - salinity slope would have changed very little between present-day (slope of 0.21, cf Figure III.22a) and glacial climates.

Spatial vs temporal slopes

That the slope may hold under LGM climate is not sufficient for reconstructing paleosalinity . Instead of the spatial slope, a temporal slope must be used, i.e. the slope relating the time variations of d¹⁸O and salinity. Figure III.22b illustrates this difference. We denote

D

d_glob and

D

S_glob the global variations of d¹⁸O and salinity due to the ice cap volume, both close to +1ä for the present-day to LGM variations (the actual values are not important for the diagram). This oceanic enrichment shifts the present-day linear relationship onto a parallel line.

The intercept I is increased, not by

D

d_glob alone but actually by

D

d_glob -

a

D

S_glob , with

a

the slope of the relationship (thus the increase is about 0.8 ä on Figure III.22b). The residuals

D

d and

D

S between these global homogeneous variations and the actual ones (

D

d = d¹⁸O_LGM- d¹⁸O_modern-

D

d_glob and

D

S = S_LGM - S_modern -

D

S_glob ) are due to local hydrological and/or oceanic circulation changes. When used to reconstruct past salinity variations, these residuals are supposed to be related by the present-day spatial slope

a

, that is

D

S =

D

a

. If not true, i.e. if the temporal slope relating

D

d and

D

S is not equal to

a

or not constant over the basin, the slope of the LGM relationship would be different to

a

and the intercept different to

I + D

d_glob -

a

D

S_glob . Indeed reconstruction discussed above predicts an intercept higher than the assumed one (-6.2ä, Figure III.22b). Given the different spatial slope

aÕ

due to this difference of intercept, Equation 3 allows calculation of the deviation

e

to the assumed salinity variation

D

S =

D

/a

a * a©

modern

*

D

Using the values simulated by the box model and discussed above (

a

=0.21,

aÕ

=0.15 to 0.20, d₀=0.2ä) , and for typical d_modern values around 0.2ä (Figure III.19), a local variation of

D

d = 0.3ä would imply a local salinity variation

D

S =

D

/a=

1.45ä , under-estimated by

e

= 0.6 to 0.1ä (depending on the I_LGM value).

An estimation of the uncertainty due to the complete inference procedure can be found in Schmidt (1999).

No documento isotopique des précipitations antarctiques : simulation pour différents climats (páginas 136-142)